Inclusion in Druntime - Githubissues

CyberShadow commented 9 years ago

Hi, just to let you know - there are two efforts underway to integrate these bindings into Druntime, so they can be distributed with D:

https://github.com/D-Programming-Language/druntime/pull/1076 https://github.com/D-Programming-Language/druntime/pull/1402

Hopefully once one of these is merged, third-party repositories will become unnecessary and development can continue in D's GitHub repo.

CyberShadow commented 8 years ago

The bindings are now part of the 2.070 D release: http://dlang.org/changelog/2.070.0.html

Manuzor commented 8 years ago

@CyberShadow You got me pretty excited there for a minute :smile: But I don't see any of the DirectX, XAudio, or XInput bindings in 2.070.2, unfortunately. It sounded like you meant that all these bindings were adopted.

CyberShadow commented 8 years ago

I decided to leave out bindings that did not clearly have license information that permitted inclusion (in this case, all bindings which did not have a public domain header). That included the Winsock2 and all DirectX modules.

marler8997 commented 8 years ago

I'm not sure that porting a header file to another language and distributing it would violate any sort of license. I'd like to see more of the bindings integrated to druntime. Is there anything I could do to help with this? I could look into the licensing if that would help? I work at HP and we use windows CE and XP on our printers so there's quite a few people here who are familiar with these types of details I could talk to.

Manuzor commented 8 years ago

I have since written a tool that converts MS .idl files to D which I used to generate D3D11 bindings. Doing manual conversion seemed kind of insane to me since there are so many what with all the recent revisions of / additions to D3D11. Unfortunately, that tool does not 100% automate this process. Because MS IDL allows for raw c++ code insertion using the cpp_quote command (yay...) my conversion tool does not touch them and just copies them over verbatim, adding a static_assert that they still need manual conversion.

The results of that tool can be found here: krepel-d/code/directx. For reference, this folder also contains the original .idl files.

The resulting .d files don't have a license yet or anything (because I didn't care so far) but I'd gladly share those!

Manuzor commented 8 years ago

If you're interested, I've moved the conversion tool here.

CyberShadow commented 8 years ago

Honestly there isn't much left.

I'm fairly certain that the DirectX bindings don't belong in Druntime proper, since nothing in Phobos or Druntime needs them. They would be a good candidate for Deimos.

That just leaves the WinSock bindings, at which point it'd be easier to simply improve the existing ones.

smjgordon commented 8 years ago

@marler8997 If a header is translated from one language to another, the translated header is a derivative work of the original header, and therefore the licence of the original header is relevant. It depends on what that licence says about derivative works.

marler8997 commented 8 years ago

@smjgordon yes I suppose that makes sense. Do we know how the headers are licensed? I would imagine it would have to be a very open license (for the headers).

smjgordon commented 8 years ago

@marler8997 The MinGW headers were public domain when we started the project. However, this seems to have changed to a MIT-style licence at some point. I'm not sure what that means. I guess that once you've released something into the public domain you can't really take it out again, and so the older versions would still be PD, but the newer updates to the files are MIT licensed.

CyberShadow commented 8 years ago

FWIW:

winsock2.d has an explicit copyright notice, but no license information. Generally, by default this means that any redistribution is prohibited.
Public domain is not recognized in all jurisdictions, which is why it is generally frowned upon to release software into the public domain - an extremely permissive license is instead preferred. This is probably why the license was changed to MIT.
Public domain means that it is not copyrighted by anyone, which means that anyone can claim a copy as their own and release it under an arbitrary license. This is probably what allowed the license to be changed to MIT. I have also taken advantage of this to relicense the Druntime versions under the Boost software license, for the reason mentioned above and for consistency with the rest of Phobos and Druntime.

marler8997 commented 8 years ago

So I guess that leaves us with 2 options for winsock in druntime:

Modify what's in druntime right now based on the windows documentation (instead of the headers themselves) so it's not a "derivative work" of the headers.
Use a derivative work of the MinGW winsock2 headers (MIT is a very open license, if I remember correctly the only requirement is to include the MIT license in any derivative work and maintain the copyright notice).

Does this sound correct?

It's a bit silly that we can't use what we already have, but this sort of thing does happen when it comes to licensing. We would probably be OK using the code based of the winsock2 headers, but these 2 options are pretty full proof IMO.

CyberShadow commented 8 years ago

Use a derivative work of the MinGW winsock2 headers (MIT is a very open license, if I remember correctly the only requirement is to include the MIT license in any derivative work and maintain the copyright notice).

Yes, and this is an additional burden that Phobos/Druntime currently does not have. Depending on how you interpret it, it would mean that a copy of the MIT license would need to be redistributed with ANY Windows program compiled with a D compiler, which is obviously unreasonable for a programming language.

Modify what's in druntime right now based on the windows documentation (instead of the headers themselves) so it's not a "derivative work" of the headers.

This is the way to go. Another correct way to do it, when applicable, is to write small C programs which print the results of including Windows headers, such as struct sizes or offsets, and use these numbers to implement the D versions.

marler8997 commented 8 years ago

Use a derivative work of the MinGW winsock2 headers (MIT is a very open license, if I remember correctly the only requirement is to include the MIT license in any derivative work and maintain the copyright notice).

Yes, and this is an additional burden that Phobos/Druntime currently does not have. Depending on how you interpret it, it would mean that a copy of the MIT license would need to be redistributed with ANY Windows program compiled with a D compiler, which is obviously unreasonable for a programming language.

Yes, requiring every application that uses the druntime to include the MIT license would be VERY bad. Well it looks like we have a clear path forward. I could work on this and submit a PR but I'd want to make sure no one else is currently working on it and that this code is something that people want included in druntime. But if someone else wants to do this I'm fine with that as well. How should we move forward?

CyberShadow commented 8 years ago

but I'd want to make sure no one else is currently working on it

As far as I know, the only person who might be already working on it is @qchikara. He recently submitted a huge patch that fixes a lot of declarations from the adapted headers:

https://github.com/dlang/druntime/pull/1576

Perhaps you could coordinate with him. Otherwise I think it's unlikely that someone is already working on this particular issue.

qchikara commented 8 years ago

As far as I know, the only person who might be already working on it is @qchikara. He recently submitted a huge patch that fixes a lot of declarations from the adapted headers:

Hi. I've already finished my work and am not going to modify this. So feel free to do your works, please.

smjgordon / bindings

Inclusion in Druntime #4